Handwriting-synthesis

by swechhachoudhary

swechhachoudhary / Handwriting-synthesis

Implementation of "Generating Sequences With Recurrent Neural Networks" https://arxiv.org/abs/1308.0...

129 Stars 12 Forks Last release: Not found MIT License 149 Commits 0 Releases

Available items

No Items, yet!

The developer of this repository has not created any items for sale yet. Need a bug fixed? Help with integration? A different license? Create a request here:

Handwriting generation

This project is a web app (CopyMonkey) which uses machine learning to mimic your handwriting style like a monkey. Have fun generating handwriting in your own style!

Demo: https://youtu.be/Ghsb3w0QACI

The ML model is implemented from "Generating Sequences With Recurrent Neural Networks" by Alex Graves (https://arxiv.org/abs/1308.0850)

How to run the web app on your local machine

git clone https://github.com/swechhachoudhary/Handwriting-synthesis.git
cd Handwriting-synthesis
python3 -m venv hand_gen_env
source hand_gen_env/bin/activate
pip install -r requirements.txt
python main.py

How to train handwriting generation (synthesis) model

python train.py --n_epochs 120 --model synthesis --batch_size 64 --text_req --n_epochs 200

How to train handwriting prediction model

python train.py --n_epochs 120 --model prediction --batch_size 64 --n_epochs 200

Data description:

There are 2 data files that you need to consider:

data.npy
and
sentences.txt
.
data.npy
contains 6000 sequences of points that correspond to handwritten sentences.
sentences.txt
contains the corresponding text sentences. You can see an example on how to load and plot an example sentence in
example.ipynb
. Each handwritten sentence is represented as a 2D array with T rows and 3 columns. T is the number of timesteps. The first column represents whether to interrupt the current stroke (i.e. when the pen is lifted off the paper). The second and third columns represent the relative coordinates of the new point with respect to the last point. Please have a look at the plot_stroke if you want to understand how to plot this sequence.

Unconditional generation.

Generated samples: alt text alt text Samples generated using priming: * Prime style text is "medical assistance", text after this is generated by model alt text * Prime style text is "something which he is passing on", text after this is generated by model alt text * Prime style text is "In Africa Jones hotels spring", text after this is generated by model alt text

Conditional generation.

Generated samples: alt text alt text alt text

We use cookies. If you continue to browse the site, you agree to the use of cookies. For more information on our use of cookies please see our Privacy Policy.